Error on Xenmotion when XS host has to perform memory ballooning

Under moderate to heavy migration loads, I'm getting Xenmotion errors now on our XenServer pool that handles XenDesktop 7.7 VMs when evacuating a XenServer or manually performing Xenmotion on one or more VMs. I just patched up through and including hotfix XS65ESP1034 and have never seen this before. This failure seems to be taking place only when memory ballooning is underway (either growing or shrinking VM memory allocation) is taking place is my conclusion from the latest tests. Although xenopsd complains about the lack of memory, there is still plenty of sufficient capacity for the server to squeeze/unsqueeze VMs to create.free memory (several tens of GB) even though all physical memory on the XS is allocated. Subsequent attempts to relaunch the evacuation or move a VM vary as to success rate. Eventually, it is possible to get all the Xenmotions performed with persistence, but experiencing multiple failures during the process.

There were around 100 - 120 VMs running on the affected XS hosts, with a VM memory range of 2 - 3 GB assigned to the majority of the running VMs.

Once pulling a VM from or pushing a VM to a XS with sufficient memory such that all VMs have expanded back to their maximum memory allocation, the issue disappears. Not sure if this also takes place on XS 7.0 or other XenServer versions, as I have no comparable such environment.